When I started in financial modeling some time ago, I had the luck to work with some top level quants. One comment that I still recall from these times is that all complex modeling problems should be reduced to two simple questions: what is my most likely outcome and how big is its variability. Very same statement can be made about (re)insurance loss modeling. Though particularly in our industry, more and more there is the reverse question being asked of when knowing that paying all (re)insured limits will lead to 'utter ruin' - to use an economics term - what is a fair probability (weight) estimate of such an fully ruinous outcome. Is this really the reverse question of the most likely outcome - the former being a quantity problem and the latter being a likelihood problem, or is it a question subject to a different extreme and rare event - scenario modeling regime. There is certainly both an academic and practitioners' school of thought which supports that the modeling of rare and extreme catastrophe scenarios is a separate and distinguishable modeling regime from that of finding the most likely outcome of a stochastic loss simulation. There is agreement in one subject regardless - dependencies, conditionality and correlations have a much stronger marginal impacts in rare and extreme scenarios than in most likely outcomes. What makes this academic research significantly relevant to underwriting is strict attention payed to accumulation of exposed limits - i.e. the summation of all (re)insured limits written by the firm. Adding a probability weight, which describes the likelihood of this outcome in context of a geo-temporal modeling regime makes that summation monetary amount that much more accurate. And with time, I have learned that the availability of physical good quality data is the most challenging aspect of developing tail modeling regimes (some details on their statistical mechanics in this earlier study: http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2341678). Lastly, to end on a more positive note, the accumulation and storage aspect of historical data is now being advanced by the promise of big data technologies.
Category: Climate/natural disasters: Climate change, Disaster risk, Floods/storms